Currently each blkcg_gq holds a request_queue reference, which is what
is used in the policies. But a lot of these interfaces will move over to
use a gendisk, so store a disk in struct blkcg_gq and hold a reference to
it.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Andreas Herrmann <aherrmann@suse.de>
Acked-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/r/20230203150400.3199230-7-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The main service scheme of BFQ for sync I/O is serving one sync
bfq_queue at a time, for a while. In particular, BFQ enforces this
scheme when it deems the latter necessary to boost throughput or
to preserve service guarantees. Unfortunately, when BFQ enforces
this policy, only one actuator at a time gets served for a while,
because each bfq_queue contains I/O only for one actuator. The
other actuators may remain underutilized.
Actually, BFQ may serve (inject) extra I/O, taken from other
bfq_queues, in parallel with that of the in-service queue. This
injection mechanism may provide the ground for dealing also with
the above actuator-underutilization problem. Yet BFQ does not take
the actuator load into account when choosing which queue to pick
extra I/O from. In addition, BFQ may happen to inject extra I/O
only when the in-service queue is temporarily empty.
In view of these facts, this commit extends the
injection mechanism in such a way that the latter:
(1) takes into account also the actuator load;
(2) checks such a load on each dispatch, and injects I/O for an
underutilized actuator, if there is one and there is I/O for it.
To perform the check in (2), this commit introduces a load
threshold, currently set to 4. A linear scan of each actuator is
performed, until an actuator is found for which the following two
conditions hold: the load of the actuator is below the threshold,
and there is at least one non-in-service queue that contains I/O
for that actuator. If such a pair (actuator, queue) is found, then
the head request of that queue is returned for dispatch, instead
of the head request of the in-service queue.
We have set the threshold, empirically, to the minimum possible
value for which an actuator is fully utilized, or close to be
fully utilized. By doing so, injected I/O 'steals' as few
drive-queue slots as possibile to the in-service queue. This
reduces as much as possible the probability that the service of
I/O from the in-service bfq_queue gets delayed because of slot
exhaustion, i.e., because all the slots of the drive queue are
filled with I/O injected from other queues (NCQ provides for 32
slots).
This new mechanism also counters actuator underutilization in the
case of asymmetric configurations of bfq_queues. Namely if there
are few bfq_queues containing I/O for some actuators and many
bfq_queues containing I/O for other actuators. Or if the
bfq_queues containing I/O for some actuators have lower weights
than the other bfq_queues.
Reviewed-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Davide Zini <davidezini2@gmail.com>
Link: https://lore.kernel.org/r/20230103145503.71712-8-paolo.valente@linaro.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Single-LUN multi-actuator SCSI drives, as well as all multi-actuator
SATA drives appear as a single device to the I/O subsystem [1]. Yet
they address commands to different actuators internally, as a function
of Logical Block Addressing (LBAs). A given sector is reachable by
only one of the actuators. For example, Seagate’s Serial Advanced
Technology Attachment (SATA) version contains two actuators and maps
the lower half of the SATA LBA space to the lower actuator and the
upper half to the upper actuator.
Evidently, to fully utilize actuators, no actuator must be left idle
or underutilized while there is pending I/O for it. The block layer
must somehow control the load of each actuator individually. This
commit lays the ground for allowing BFQ to provide such a per-actuator
control.
BFQ associates an I/O-request sync bfq_queue with each process doing
synchronous I/O, or with a group of processes, in case of queue
merging. Then BFQ serves one bfq_queue at a time. While in service, a
bfq_queue is emptied in request-position order. Yet the same process,
or group of processes, may generate I/O for different actuators. In
this case, different streams of I/O (each for a different actuator)
get all inserted into the same sync bfq_queue. So there is basically
no individual control on when each stream is served, i.e., on when the
I/O requests of the stream are picked from the bfq_queue and
dispatched to the drive.
This commit enables BFQ to control the service of each actuator
individually for synchronous I/O, by simply splitting each sync
bfq_queue into N queues, one for each actuator. In other words, a sync
bfq_queue is now associated to a pair (process, actuator). As a
consequence of this split, the per-queue proportional-share policy
implemented by BFQ will guarantee that the sync I/O generated for each
actuator, by each process, receives its fair share of service.
This is just a preparatory patch. If the I/O of the same process
happens to be sent to different queues, then each of these queues may
undergo queue merging. To handle this event, the bfq_io_cq data
structure must be properly extended. In addition, stable merging must
be disabled to avoid loss of control on individual actuators. Finally,
also async queues must be split. These issues are described in detail
and addressed in next commits. As for this commit, although multiple
per-process bfq_queues are provided, the I/O of each process or group
of processes is still sent to only one queue, regardless of the
actuator the I/O is for. The forwarding to distinct bfq_queues will be
enabled after addressing the above issues.
[1] https://www.linaro.org/blog/budget-fair-queueing-bfq-linux-io-scheduler-optimizations-for-multi-actuator-sata-hard-drives/
Reviewed-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Signed-off-by: Gabriele Felici <felicigb@gmail.com>
Signed-off-by: Carmine Zaccagnino <carmine@carminezacc.com>
Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Link: https://lore.kernel.org/r/20230103145503.71712-2-paolo.valente@linaro.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The updating of 'bfqg->ref' should be protected by 'bfqd->lock', however,
during code review, we found that bfq_pd_free() update 'bfqg->ref'
without holding the lock, which is problematic:
1) bfq_pd_free() triggered by removing cgroup is called asynchronously;
2) bfqq will grab bfqg reference, and exit bfqq will drop the reference,
which can concurrent with 1).
Unfortunately, 'bfqd->lock' can't be held here because 'bfqd' might already
be freed in bfq_pd_free(). Fix the problem by using atomic refcount apis.
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230103084755.1256479-1-yukuai1@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
-----BEGIN PGP SIGNATURE-----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=k/vv
-----END PGP SIGNATURE-----
Merge tag 'for-6.2/block-2022-12-08' of git://git.kernel.dk/linux
Pull block updates from Jens Axboe:
- NVMe pull requests via Christoph:
- Support some passthrough commands without CAP_SYS_ADMIN (Kanchan
Joshi)
- Refactor PCIe probing and reset (Christoph Hellwig)
- Various fabrics authentication fixes and improvements (Sagi
Grimberg)
- Avoid fallback to sequential scan due to transient issues (Uday
Shankar)
- Implement support for the DEAC bit in Write Zeroes (Christoph
Hellwig)
- Allow overriding the IEEE OUI and firmware revision in configfs
for nvmet (Aleksandr Miloserdov)
- Force reconnect when number of queue changes in nvmet (Daniel
Wagner)
- Minor fixes and improvements (Uros Bizjak, Joel Granados, Sagi
Grimberg, Christoph Hellwig, Christophe JAILLET)
- Fix and cleanup nvme-fc req allocation (Chaitanya Kulkarni)
- Use the common tagset helpers in nvme-pci driver (Christoph
Hellwig)
- Cleanup the nvme-pci removal path (Christoph Hellwig)
- Use kstrtobool() instead of strtobool (Christophe JAILLET)
- Allow unprivileged passthrough of Identify Controller (Joel
Granados)
- Support io stats on the mpath device (Sagi Grimberg)
- Minor nvmet cleanup (Sagi Grimberg)
- MD pull requests via Song:
- Code cleanups (Christoph)
- Various fixes
- Floppy pull request from Denis:
- Fix a memory leak in the init error path (Yuan)
- Series fixing some batch wakeup issues with sbitmap (Gabriel)
- Removal of the pktcdvd driver that was deprecated more than 5 years
ago, and subsequent removal of the devnode callback in struct
block_device_operations as no users are now left (Greg)
- Fix for partition read on an exclusively opened bdev (Jan)
- Series of elevator API cleanups (Jinlong, Christoph)
- Series of fixes and cleanups for blk-iocost (Kemeng)
- Series of fixes and cleanups for blk-throttle (Kemeng)
- Series adding concurrent support for sync queues in BFQ (Yu)
- Series bringing drbd a bit closer to the out-of-tree maintained
version (Christian, Joel, Lars, Philipp)
- Misc drbd fixes (Wang)
- blk-wbt fixes and tweaks for enable/disable (Yu)
- Fixes for mq-deadline for zoned devices (Damien)
- Add support for read-only and offline zones for null_blk
(Shin'ichiro)
- Series fixing the delayed holder tracking, as used by DM (Yu,
Christoph)
- Series enabling bio alloc caching for IRQ based IO (Pavel)
- Series enabling userspace peer-to-peer DMA (Logan)
- BFQ waker fixes (Khazhismel)
- Series fixing elevator refcount issues (Christoph, Jinlong)
- Series cleaning up references around queue destruction (Christoph)
- Series doing quiesce by tagset, enabling cleanups in drivers
(Christoph, Chao)
- Series untangling the queue kobject and queue references (Christoph)
- Misc fixes and cleanups (Bart, David, Dawei, Jinlong, Kemeng, Ye,
Yang, Waiman, Shin'ichiro, Randy, Pankaj, Christoph)
* tag 'for-6.2/block-2022-12-08' of git://git.kernel.dk/linux: (247 commits)
blktrace: Fix output non-blktrace event when blk_classic option enabled
block: sed-opal: Don't include <linux/kernel.h>
sed-opal: allow using IOC_OPAL_SAVE for locking too
blk-cgroup: Fix typo in comment
block: remove bio_set_op_attrs
nvmet: don't open-code NVME_NS_ATTR_RO enumeration
nvme-pci: use the tagset alloc/free helpers
nvme: add the Apple shared tag workaround to nvme_alloc_io_tag_set
nvme: only set reserved_tags in nvme_alloc_io_tag_set for fabrics controllers
nvme: consolidate setting the tagset flags
nvme: pass nr_maps explicitly to nvme_alloc_io_tag_set
block: bio_copy_data_iter
nvme-pci: split out a nvme_pci_ctrl_is_dead helper
nvme-pci: return early on ctrl state mismatch in nvme_reset_work
nvme-pci: rename nvme_disable_io_queues
nvme-pci: cleanup nvme_suspend_queue
nvme-pci: remove nvme_pci_disable
nvme-pci: remove nvme_disable_admin_queue
nvme: merge nvme_shutdown_ctrl into nvme_disable_ctrl
nvme: use nvme_wait_ready in nvme_shutdown_ctrl
...
Prevent unnecessary format conversion for bfqg->bfqd in multiple
places.
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Acked-by: Paolo Valente <paolo.valente@unimore.it>
Link: https://lore.kernel.org/r/20221102022542.3621219-6-yukuai1@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Prepare to refactor the counting of 'num_groups_with_pending_reqs'.
Add a counter in bfq_group, update it while tracking if bfqq have pending
requests and when bfq_bfqq_move() is called.
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Acked-by: Paolo Valente <paolo.valente@linaro.org>
Link: https://lore.kernel.org/r/20220916071942.214222-3-yukuai1@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
While doing code coverage testing(CONFIG_BFQ_CGROUP_DEBUG is disabled), we
found that some functions doesn't have caller, thus remove them.
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20220816015631.1323948-2-yukuai1@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Use the new blk_opf_t type for arguments and variables that represent
request flags or a bitwise combination of a request operation and
request flags. Rename those variables from 'op' into 'opf'.
This patch does not change any functionality.
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20220714180729.1065367-8-bvanassche@acm.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Fix the following warnings:
block/bfq-cgroup.c:721: warning: Function parameter or member 'bfqg' not described in '__bfq_bic_change_cgroup'
block/bfq-cgroup.c:721: warning: Excess function parameter 'blkcg' description in '__bfq_bic_change_cgroup'
block/bfq-cgroup.c:870: warning: Function parameter or member 'ioprio_class' not described in 'bfq_reparent_leaf_entity'
block/bfq-cgroup.c:900: warning: Function parameter or member 'ioprio_class' not described in 'bfq_reparent_active_queues'
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Link: https://lore.kernel.org/r/20220617210859.106623-1-bvanassche@acm.org
Bios queued into BFQ IO scheduler can be associated with a cgroup that
was already offlined. This may then cause insertion of this bfq_group
into a service tree. But this bfq_group will get freed as soon as last
bio associated with it is completed leading to use after free issues for
service tree users. Fix the problem by making sure we always operate on
online bfq_group. If the bfq_group associated with the bio is not
online, we pick the first online parent.
CC: stable@vger.kernel.org
Fixes: e21b7a0b98 ("block, bfq: add full hierarchical scheduling and cgroups support")
Tested-by: "yukuai (C)" <yukuai3@huawei.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220401102752.8599-9-jack@suse.cz
Signed-off-by: Jens Axboe <axboe@kernel.dk>
BFQ usage of __bio_blkcg() is a relict from the past. Furthermore if bio
would not be associated with any blkcg, the usage of __bio_blkcg() in
BFQ is prone to races with the task being migrated between cgroups as
__bio_blkcg() calls at different places could return different blkcgs.
Convert BFQ to the new situation where bio->bi_blkg is initialized in
bio_set_dev() and thus practically always valid. This allows us to save
blkcg_gq lookup and noticeably simplify the code.
CC: stable@vger.kernel.org
Fixes: 0fe061b9f0 ("blkcg: fix ref count issue with bio_blkcg() using task_css")
Tested-by: "yukuai (C)" <yukuai3@huawei.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220401102752.8599-8-jack@suse.cz
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Track whether bfq_group is still online. We cannot rely on
blkcg_gq->online because that gets cleared only after all policies are
offlined and we need something that gets updated already under
bfqd->lock when we are cleaning up our bfq_group to be able to guarantee
that when we see online bfq_group, it will stay online while we are
holding bfqd->lock lock.
CC: stable@vger.kernel.org
Tested-by: "yukuai (C)" <yukuai3@huawei.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220401102752.8599-7-jack@suse.cz
Signed-off-by: Jens Axboe <axboe@kernel.dk>
When bfqq is shared by multiple processes it can happen that one of the
processes gets moved to a different cgroup (or just starts submitting IO
for different cgroup). In case that happens we need to split the merged
bfqq as otherwise we will have IO for multiple cgroups in one bfqq and
we will just account IO time to wrong entities etc.
Similarly if the bfqq is scheduled to merge with another bfqq but the
merge didn't happen yet, cancel the merge as it need not be valid
anymore.
CC: stable@vger.kernel.org
Fixes: e21b7a0b98 ("block, bfq: add full hierarchical scheduling and cgroups support")
Tested-by: "yukuai (C)" <yukuai3@huawei.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20220401102752.8599-3-jack@suse.cz
Signed-off-by: Jens Axboe <axboe@kernel.dk>
In bfq_pd_alloc(), the function bfqg_stats_init() init bfqg. If
blkg_rwstat_init() init bfqg_stats->bytes successful and init
bfqg_stats->ios failed, bfqg_stats_init() return failed, bfqg will
be freed. But blkg_rwstat->cpu_cnt is not deleted from the list of
percpu_counters. If we traverse the list of percpu_counters, It will
have UAF problem.
we should use blkg_rwstat_exit() to cleanup bfqg_stats bytes in the
above scenario.
Fixes: commit fd41e60331 ("bfq-iosched: stop using blkg->stat_bytes and ->stat_ios")
Signed-off-by: Zheng Liang <zhengliang6@huawei.com>
Acked-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/r/20211018024225.1493938-1-zhengliang6@huawei.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Except for the features passed to blk_queue_required_elevator_features,
elevator.h is only needed internally to the block layer. Move the
ELEVATOR_F_* definitions to blkdev.h, and the move elevator.h to
block/, dropping all the spurious includes outside of that.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Link: https://lore.kernel.org/r/20210920123328.1399408-13-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Since commit 430a67f9d6 ("block, bfq: merge bursts of newly-created
queues"), BFQ maintains a per-group pointer to the last bfq_queue
created. If such a queue, say bfqq, happens to move to a different
group, then bfqq is no more a valid last bfq_queue created for its
previous group. That pointer must then be cleared. Not resetting such
a pointer may also cause UAF, if bfqq happens to also be freed after
being moved to a different group. This commit performs this missing
reset. As such it fixes commit 430a67f9d6 ("block, bfq: merge bursts
of newly-created queues").
Such a missing reset is most likely the cause of the crash reported in [1].
With some analysis, we found that this crash was due to the
above UAF. And such UAF did go away with this commit applied [1].
Anyway, before this commit, that crash happened to be triggered in
conjunction with commit 2d52c58b9c ("block, bfq: honor already-setup
queue merges"). The latter was then reverted by commit ebc69e897e
("Revert "block, bfq: honor already-setup queue merges""). Yet commit
2d52c58b9c ("block, bfq: honor already-setup queue merges") contains
no error related with the above UAF, and can then be restored.
[1] https://bugzilla.kernel.org/show_bug.cgi?id=214503
Fixes: 430a67f9d6 ("block, bfq: merge bursts of newly-created queues")
Tested-by: Grzegorz Kowal <custos.mentis@gmail.com>
Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Link: https://lore.kernel.org/r/20211015144336.45894-2-paolo.valente@linaro.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Many throughput-sensitive workloads are made of several parallel I/O
flows, with all flows generated by the same application, or more
generically by the same task (e.g., system boot). The most
counterproductive action with these workloads is plugging I/O dispatch
when one of the bfq_queues associated with these flows remains
temporarily empty.
To avoid this plugging, BFQ has been using a burst-handling mechanism
for years now. This mechanism has proven effective for throughput, and
not detrimental for service guarantees. This commit pushes this
mechanism a little bit further, basing on the following two facts.
First, all the I/O flows of a the same application or task contribute
to the execution/completion of that common application or task. So the
performance figures that matter are total throughput of the flows and
task-wide I/O latency. In particular, these flows do not need to be
protected from each other, in terms of individual bandwidth or
latency.
Second, the above fact holds regardless of the number of flows.
Putting these two facts together, this commits merges stably the
bfq_queues associated with these I/O flows, i.e., with the processes
that generate these IO/ flows, regardless of how many the involved
processes are.
To decide whether a set of bfq_queues is actually associated with the
I/O flows of a common application or task, and to merge these queues
stably, this commit operates as follows: given a bfq_queue, say Q2,
currently being created, and the last bfq_queue, say Q1, created
before Q2, Q2 is merged stably with Q1 if
- very little time has elapsed since when Q1 was created
- Q2 has the same ioprio as Q1
- Q2 belongs to the same group as Q1
Merging bfq_queues also reduces scheduling overhead. A fio test with
ten random readers on /dev/nullb shows a throughput boost of 40%, with
a quadcore. Since BFQ's execution time amounts to ~50% of the total
per-request processing time, the above throughput boost implies that
BFQ's overhead is reduced by more than 50%.
Tested-by: Jan Kara <jack@suse.cz>
Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Tested-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Link: https://lore.kernel.org/r/20210304174627.161-7-paolo.valente@linaro.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Changes from v1:
- update commit description with proper ref-accounting justification
commit db37a34c56 ("block, bfq: get a ref to a group when adding it to a service tree")
introduce leak forbfq_group and blkcg_gq objects because of get/put
imbalance.
In fact whole idea of original commit is wrong because bfq_group entity
can not dissapear under us because it is referenced by child bfq_queue's
entities from here:
-> bfq_init_entity()
->bfqg_and_blkg_get(bfqg);
->entity->parent = bfqg->my_entity
-> bfq_put_queue(bfqq)
FINAL_PUT
->bfqg_and_blkg_put(bfqq_group(bfqq))
->kmem_cache_free(bfq_pool, bfqq);
So parent entity can not disappear while child entity is in tree,
and child entities already has proper protection.
This patch revert commit db37a34c56 ("block, bfq: get a ref to a group when adding it to a service tree")
bfq_group leak trace caused by bad commit:
-> blkg_alloc
-> bfq_pq_alloc
-> bfqg_get (+1)
->bfq_activate_bfqq
->bfq_activate_requeue_entity
-> __bfq_activate_entity
->bfq_get_entity
->bfqg_and_blkg_get (+1) <==== : Note1
->bfq_del_bfqq_busy
->bfq_deactivate_entity+0x53/0xc0 [bfq]
->__bfq_deactivate_entity+0x1b8/0x210 [bfq]
-> bfq_forget_entity(is_in_service = true)
entity->on_st_or_in_serv = false <=== :Note2
if (is_in_service)
return; ==> do not touch reference
-> blkcg_css_offline
-> blkcg_destroy_blkgs
-> blkg_destroy
-> bfq_pd_offline
-> __bfq_deactivate_entity
if (!entity->on_st_or_in_serv) /* true, because (Note2)
return false;
-> bfq_pd_free
-> bfqg_put() (-1, byt bfqg->ref == 2) because of (Note2)
So bfq_group and blkcg_gq will leak forever, see test-case below.
##TESTCASE_BEGIN:
#!/bin/bash
max_iters=${1:-100}
#prep cgroup mounts
mount -t tmpfs cgroup_root /sys/fs/cgroup
mkdir /sys/fs/cgroup/blkio
mount -t cgroup -o blkio none /sys/fs/cgroup/blkio
# Prepare blkdev
grep blkio /proc/cgroups
truncate -s 1M img
losetup /dev/loop0 img
echo bfq > /sys/block/loop0/queue/scheduler
grep blkio /proc/cgroups
for ((i=0;i<max_iters;i++))
do
mkdir -p /sys/fs/cgroup/blkio/a
echo 0 > /sys/fs/cgroup/blkio/a/cgroup.procs
dd if=/dev/loop0 bs=4k count=1 of=/dev/null iflag=direct 2> /dev/null
echo 0 > /sys/fs/cgroup/blkio/cgroup.procs
rmdir /sys/fs/cgroup/blkio/a
grep blkio /proc/cgroups
done
##TESTCASE_END:
Fixes: db37a34c56 ("block, bfq: get a ref to a group when adding it to a service tree")
Tested-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Signed-off-by: Dmitry Monakhov <dmtrmonakhov@yandex-team.ru>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
In bfq_pd_offline(), the function bfq_flush_idle_tree() is invoked to
flush the rb tree that contains all idle entities belonging to the pd
(cgroup) being destroyed. In particular, bfq_flush_idle_tree() is
invoked before bfq_reparent_active_queues(). Yet the latter may happen
to add some entities to the idle tree. It happens if, in some of the
calls to bfq_bfqq_move() performed by bfq_reparent_active_queues(),
the queue to move is empty and gets expired.
This commit simply reverses the invocation order between
bfq_flush_idle_tree() and bfq_reparent_active_queues().
Tested-by: cki-project@redhat.com
Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
bfq_reparent_leaf_entity() reparents the input leaf entity (a leaf
entity represents just a bfq_queue in an entity tree). Yet, the input
entity is guaranteed to always be a leaf entity only in two-level
entity trees. In this respect, because of the error fixed by
commit 14afc59361 ("block, bfq: fix overwrite of bfq_group pointer
in bfq_find_set_group()"), all (wrongly collapsed) entity trees happened
to actually have only two levels. After the latter commit, this does not
hold any longer.
This commit fixes this problem by modifying
bfq_reparent_leaf_entity(), so that it searches an active leaf entity
down the path that stems from the input entity. Such a leaf entity is
guaranteed to exist when bfq_reparent_leaf_entity() is invoked.
Tested-by: cki-project@redhat.com
Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
A bfq_put_queue() may be invoked in __bfq_bic_change_cgroup(). The
goal of this put is to release a process reference to a bfq_queue. But
process-reference releases may trigger also some extra operation, and,
to this goal, are handled through bfq_release_process_ref(). So, turn
the invocation of bfq_put_queue() into an invocation of
bfq_release_process_ref().
Tested-by: cki-project@redhat.com
Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Commit ecedd3d7e1 ("block, bfq: get extra ref to prevent a queue
from being freed during a group move") gets an extra reference to a
bfq_queue before possibly deactivating it (temporarily), in
bfq_bfqq_move(). This prevents the bfq_queue from disappearing before
being reactivated in its new group.
Yet, the bfq_queue may also be expired (i.e., its service may be
stopped) before the bfq_queue is deactivated. And also an expiration
may lead to a premature freeing. This commit fixes this issue by
simply moving forward the getting of the extra reference already
introduced by commit ecedd3d7e1 ("block, bfq: get extra ref to
prevent a queue from being freed during a group move").
Reported-by: cki-project@redhat.com
Tested-by: cki-project@redhat.com
Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The bfq_find_set_group() function takes as input a blkcg (which represents
a cgroup) and retrieves the corresponding bfq_group, then it updates the
bfq internal group hierarchy (see comments inside the function for why
this is needed) and finally it returns the bfq_group.
In the hierarchy update cycle, the pointer holding the correct bfq_group
that has to be returned is mistakenly used to traverse the hierarchy
bottom to top, meaning that in each iteration it gets overwritten with the
parent of the current group. Since the update cycle stops at root's
children (depth = 2), the overwrite becomes a problem only if the blkcg
describes a cgroup at a hierarchy level deeper than that (depth > 2). In
this case the root's child that happens to be also an ancestor of the
correct bfq_group is returned. The main consequence is that processes
contained in a cgroup at depth greater than 2 are wrongly placed in the
group described above by BFQ.
This commits fixes this problem by using a different bfq_group pointer in
the update cycle in order to avoid the overwrite of the variable holding
the original group reference.
Reported-by: Kwon Je Oh <kwonje.oh2@gmail.com>
Signed-off-by: Carlo Nonato <carlo.nonato95@gmail.com>
Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
BFQ schedules generic entities, which may represent either bfq_queues
or groups of bfq_queues. When an entity is inserted into a service
tree, a reference must be taken, to make sure that the entity does not
disappear while still referred in the tree. Unfortunately, such a
reference is mistakenly taken only if the entity represents a
bfq_queue. This commit takes a reference also in case the entity
represents a group.
Tested-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Tested-by: Chris Evich <cevich@redhat.com>
Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The flag on_st in the bfq_entity data structure is true if the entity
is on a service tree or is in service. Yet the name of the field,
confusingly, does not mention the second, very important case. Extend
the name to mention the second case too.
Tested-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
In bfq_bfqq_move(), the bfq_queue, say Q, to be moved to a new group
may happen to be deactivated in the scheduling data structures of the
source group (and then activated in the destination group). If Q is
referred only by the data structures in the source group when the
deactivation happens, then Q is freed upon the deactivation.
This commit addresses this issue by getting an extra reference before
the possible deactivation, and releasing this extra reference after Q
has been moved.
Tested-by: Chris Evich <cevich@redhat.com>
Tested-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
bio->bi_blkg will be NULL when the issue of the request
has bypassed the block layer as shown in the following oops:
Internal error: Oops: 96000005 [#1] SMP
CPU: 17 PID: 2996 Comm: scsi_id Not tainted 5.4.0 #4
Call trace:
percpu_counter_add_batch+0x38/0x4c8
bfqg_stats_update_legacy_io+0x9c/0x280
bfq_insert_requests+0xbac/0x2190
blk_mq_sched_insert_request+0x288/0x670
blk_execute_rq_nowait+0x140/0x178
blk_execute_rq+0x8c/0x140
sg_io+0x604/0x9c0
scsi_cmd_ioctl+0xe38/0x10a8
scsi_cmd_blk_ioctl+0xac/0xe8
sd_ioctl+0xe4/0x238
blkdev_ioctl+0x590/0x20e0
block_ioctl+0x60/0x98
do_vfs_ioctl+0xe0/0x1b58
ksys_ioctl+0x80/0xd8
__arm64_sys_ioctl+0x40/0x78
el0_svc_handler+0xc4/0x270
so ensure its validity before using it.
Fixes: fd41e60331 ("bfq-iosched: stop using blkg->stat_bytes and ->stat_ios")
Signed-off-by: Hou Tao <houtao1@huawei.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
When used on cgroup1, bfq uses the blkg->stat_bytes and ->stat_ios
from blk-cgroup core to populate six stat knobs. blk-cgroup core is
moving away from blkg_rwstat to improve scalability and won't be able
to support this usage.
It isn't like the sharing gains all that much. Let's break it out to
dedicated rwstat counters which are updated when on cgroup1. This
makes use of bfqg_*rwstat*() helpers outside of
CONFIG_BFQ_CGROUP_DEBUG. Move them out.
v2: Compile fix when !CONFIG_BFQ_CGROUP_DEBUG.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Collect them right under #ifdef CONFIG_BFQ_CGROUP_DEBUG. The next
patch will use them from !DEBUG path and this makes it easy to move
them out of the ifdef block.
This is pure code reorganization.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This function will be useful when we update weight from the soon-coming
per-device interface.
Signed-off-by: Fam Zheng <zhengfeiran@bytedance.com>
Reviewed-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Instead of @node, pass in @q and @blkcg so that the alloc function has
more context. This doesn't cause any behavior change and will be used
by io.weight implementation.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This option is entirely bfq specific, give it an appropinquate name.
Also make it depend on CONFIG_BFQ_GROUP_IOSCHED in Kconfig, as all
the functionality already does so anyway.
Acked-by: Tejun Heo <tj@kernel.org>
Acked-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This function was moved from core block code and is way to generic.
Fold it into the only caller and simplify it based on the actually
passed arguments.
Acked-by: Tejun Heo <tj@kernel.org>
Acked-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This structure and assorted infrastructure is only used by the bfq I/O
scheduler. Move it there instead of bloating the common code.
Acked-by: Tejun Heo <tj@kernel.org>
Acked-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
When sampling the blkcg counts we don't need atomics or per-cpu
variables. Introduce a new structure just containing plain u64
counters.
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Returning a structure generates rather bad code, so switch to passing
by reference. Also don't require the structure to be zeroed and add
to the 0-initialized counters, but actually set the counters to the
calculated value.
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
There's some discussion on how to do this the best, and Tejun prefers
that BFQ just create the file itself instead of having cgroups support
a symlink feature.
Hence revert commit 54b7b868e8 and 19e9da9e86 for 5.2, and this
can be done properly for 5.3.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Many userspace tools and services use the proportional-share policy of
the blkio/io cgroups controller. The CFQ I/O scheduler implemented
this policy for the legacy block layer. To modify the weight of a
group in case CFQ was in charge, the 'weight' parameter of the group
must be modified. On the other hand, the BFQ I/O scheduler implements
the same policy in blk-mq, but, with BFQ, the parameter to modify has
a different name: bfq.weight (forced choice until legacy block was
present, because two different policies cannot share a common parameter
in cgroups).
Due to CFQ legacy, most if not all userspace configurations still use
the parameter 'weight', and for the moment do not seem likely to be
changed. But, when CFQ went away with legacy block, such a parameter
ceased to exist.
So, a simple workaround has been proposed [1] to make all
configurations work: add a symlink, named weight, to bfq.weight. This
commit adds such a symlink.
[1] https://lkml.org/lkml/2019/4/8/555
Suggested-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Angelo Ruocco <angeloruocco90@gmail.com>
Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
All these files have some form of the usual GPLv2 or later boilerplate.
Switch them to use SPDX tags instead.
Reviewed-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Some of the comments in the bfq files had typos. This patch fixes them.
Signed-off-by: Angelo Ruocco <angeloruocco90@gmail.com>
Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
To boost throughput with a set of processes doing interleaved I/O
(i.e., a set of processes whose individual I/O is random, but whose
merged cumulative I/O is sequential), BFQ merges the queues associated
with these processes, i.e., redirects the I/O of these processes into a
common, shared queue. In the shared queue, I/O requests are ordered by
their position on the medium, thus sequential I/O gets dispatched to
the device when the shared queue is served.
Queue merging costs execution time, because, to detect which queues to
merge, BFQ must maintain a list of the head I/O requests of active
queues, ordered by request positions. Measurements showed that this
costs about 10% of BFQ's total per-request processing time.
Request processing time becomes more and more critical as the speed of
the underlying storage device grows. Yet, fortunately, queue merging
is basically useless on the very devices that are so fast to make
request processing time critical. To reach a high throughput, these
devices must have many requests queued at the same time. But, in this
configuration, the internal scheduling algorithms of these devices do
also the job of queue merging: they reorder requests so as to obtain
as much as possible a sequential I/O pattern. As a consequence, with
processes doing interleaved I/O, the throughput reached by one such
device is likely to be the same, with and without queue merging.
In view of this fact, this commit disables queue merging, and all
related housekeeping, for non-rotational devices with internal
queueing. The total, single-lock-protected, per-request processing
time of BFQ drops to, e.g., 1.9 us on an Intel Core i7-2760QM@2.40GHz
(time measured with simple code instrumentation, and using the
throughput-sync.sh script of the S suite [1], in performance-profiling
mode). To put this result into context, the total,
single-lock-protected, per-request execution time of the lightest I/O
scheduler available in blk-mq, mq-deadline, is 0.7 us (mq-deadline is
~800 LOC, against ~10500 LOC for BFQ).
Disabling merging provides a further, remarkable benefit in terms of
throughput. Merging tends to make many workloads artificially more
uneven, mainly because of shared queues remaining non empty for
incomparably more time than normal queues. So, if, e.g., one of the
queues in a set of merged queues has a higher weight than a normal
queue, then the shared queue may inherit such a high weight and, by
staying almost always active, may force BFQ to perform I/O plugging
most of the time. This evidently makes it harder for BFQ to let the
device reach a high throughput.
As a practical example of this problem, and of the benefits of this
commit, we measured again the throughput in the nasty scenario
considered in previous commit messages: dbench test (in the Phoronix
suite), with 6 clients, on a filesystem with journaling, and with the
journaling daemon enjoying a higher weight than normal processes. With
this commit, the throughput grows from ~150 MB/s to ~200 MB/s on a
PLEXTOR PX-256M5 SSD. This is the same peak throughput reached by any
of the other I/O schedulers. As such, this is also likely to be the
maximum possible throughput reachable with this workload on this
device, because I/O is mostly random, and the other schedulers
basically just pass I/O requests to the drive as fast as possible.
[1] https://github.com/Algodev-github/S
Tested-by: Holger Hoffstätte <holger@applied-asynchrony.com>
Tested-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Tested-by: Francesco Pollicino <fra.fra.800@gmail.com>
Signed-off-by: Alessio Masola <alessio.masola@gmail.com>
Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
